Why Video Streaming Can be Much More Profitable Than Music Streaming (for Distributors)

Even if there are similarities for distributors in the streaming video and streaming music businesses, for most entities, if there was a choice, you’d probably choose to be in the video business, not music. 

Music streaming is good for copyright holders but pretty difficult for distributors, while video streaming is better for distributors at scale, and less favorable for copyright holders. 

Both types of streaming share core digital economics: high upfront fixed costs for content creation or licensing, followed by near-zero marginal costs for additional distribution.

But they diverge sharply in how revenue flows to copyright holders (artists/labels/studios) and distributors. 

The marginal cost of streaming additional songs is linear: play a stream, pay a fee. Video streaming is different: content is typically licensed for a flat, fixed fee covering unlimited streams.

So video streamers can reach higher margins as additional subscribers are added: the marginal cost comes mostly in the form of marketing or acquisition cost, not content rights payments. 

Music streamers, on the other hand, pay 70 percent of revenue to copyright holders, at the margin. Volume helps, but only so much. 

source: Joel Goveia 

Platforms such as Spotify pay out 70 percent of revenue to rights holders (roughly 55 percent to 60 percent to labels/masters and 10 percent to 15 percent to publishers). So distributor costs are variable with scale. 

More streams mean higher payouts.

Video streamers pay flat fees for licensing content, so digital scale economies work. 

For a video streamer, there is no per-view royalty. Netflix’s costs, for example,  are largely fixed upfront (production or licensing deals), so additional views do not increase payments to rights holders.

Video streaming licensing also means differentiation is possible. Virtually all music streamers have access to the same content.

Video licensing is restricted: content can be supplied uniquely on a single platform. Also, some video streamers (such as Netflix) can create original content and own it. 

Aspect	Similarities	Music Streaming – Copyright Holders	Music Streaming – Distributors (e.g., Spotify)	Video Streaming – Copyright Holders	Video Streaming – Distributors (e.g., Netflix)
Primary Revenue to Holders	Subscription-driven platform access	Revenue share (~55-60% masters, 10-15% publishing from platform revenue)	Pays ~70% of revenue to holders (variable)	Fixed lump-sum licensing fees or IP ownership	Fixed licensing or owns originals (no per-view royalties)
Marginal Cost Structure	Near-zero for delivery (bandwidth/storage)	Proportional to streams (recurring royalties)	Variable costs rise with usage/revenue	Upfront fixed fee (unlimited plays)	Fixed after acquisition; bandwidth only
Content Ownership	Holders license access rather than sell copies	Retain rights; perpetual licensing	No ownership; pure licensing	Retain/sell licensing rights; strong IP control	Increasingly owns originals for future value
Consumption Incentive	Scales with catalog size and user engagement	`` ++	Wants high engagement but pays more for it	Favors binge/complete viewing	Benefits from high consumption of sunk-cost content
Risk & Profit Model	High fixed costs; economies of scale at large user base	Stable but diluted per-stream payouts	Thin margins; freemium helps acquisition	Front-loaded, predictable fees	Higher margins possible; originals drive differentiation
Business Model Type	Platform economy with low marginal costs	Per-stream / pro-rata royalties	Revenue-share licensing	Flat-fee licensing or ownership	Fixed-cost licensing + owned content

The bottom line is that if an entity has a choice, it will want to be in the video streaming business rather than music streaming. 

The caveat is that some entities use music and/or video streaming as a “loss leader” to add value for some other product feature that actually drives the direct profits and profit margins. Amazon Prime video and music provide a good example.

Amazon AI Capex is a Rorschach Test

Amazon capital spending plans are a bit of a Rorschach test test these days: what the picture means tells us more about the viewer’s perceptions than the objective facts about the investments. 

figures in millions 

Investors and analysts fear the impact on free cash flow, of course. But Amazon operating margins still look decent, at the moment. 

Amazon CEO Andy Jassy's shareholder letter might put the firm’s thinking into perspective. Jassy says that although “reasonable people can disagree,” “when you identify disproportionate inflections, bet big.”

And Amazon believes AI is that sort of thing. 

“Three years after AWS launched commercially, it had a $58 million revenue run rate,” Jassy says. “Three years into this AI wave, AWS’s AI revenue run rate is over $15 billion in Q1 2026, nearly 260 times larger than AWS at that same point.”

And everyone agrees there now is excess demand for high-performance computing capability. “We still have capacity constraints that yield unserved demand,” says Jassy. As an example, he notes that ”two large AWS customers have already asked if they could buy ‘all’ of our Graviton instance capacity in 2026 (custom CPU chip).”

And it matters how big, and how fast, the new business grows. “The way AWS’s cash cycle works is that the faster AWS grows, the more short-term capex we’ll spend.” Everyone understands that. 

But that’s where the Rorschach analogy kicks in. “The free cash flow and return on invested capital  for these investments are cumulatively quite attractive a couple years after being in service,” Jassy says.

“However, in times of very high growth (like now), where the capex growth meaningfully outpaces the revenue growth, the early-years FCF is challenged until these initial tranches of capacity are being monetized and revenue growth out-paces capex growth,” Jassy says. 

What investors and analysts “see” in the ink blots is conditioned by their expectations about the size of the opportunity; the profit margins; the payback period and the likelihood that a “winner takes most” market structure will develop. 

“Every customer experience will be reinvented by AI, and there will be a slew of new experiences only possible because of AI,” Jassy notes. 

Agree or not, his views on a few key questions are emphatic”

• Is the technology over-hyped

• Are we  in “a bubble”

• Will the margins and ROIC will be appealing. 

“My strong conviction, at least for Amazon, is that the answers are no, no, and yes,” says Jassy. 

Strong opinions are held for each of those questions, and investor bets will follow. Big gains or big losses are possible. But investor decisions on where to place bets are a Rorschach test.

Will Claude Mythos Preview Help or Harm Security Suppliers?

It now seems almost routine that some new language model emerges to further disrupt some part of the computing industry. First it was chips, processors and memory. Then it was enterprise software. Now it seems to have extended to edge networks. 

The impact on security suppliers is less clear.

Claude Mythos Preview is Anthropic’s most capable frontier AI model to date, announced April 7, 2026), and seems poised to affect security software suppliers, although the direction and magnitude seem unclear. 

Many climbed on the day of the announcement, then retreated afterwards. 

Company	Pre-Announce Close (Apr 6)	Announce Day Close (Apr 7)	Latest Close (Apr 10)	% Change Announce Day (Apr 6 → 7)	% Change Since Announce (Apr 6 → 10)
CrowdStrike	$398.61	$423.23	$379.02	+6.2%	-4.9%
Palo Alto Networks	$161.95	$169.87	$155.73	+5.0%	-3.8%
Cisco	$80.44	$80.68	$82.22	+0.3%	+2.2%
Fortinet	$82.29	$83.72	$76.70	+1.7%	-6.8%
Zscaler	$139.52	$142.09	$118.05	+1.8%	-15.4%
SentinelOne	$13.51	$13.38	$11.94	-1.0%	-11.6%
Cyber Security ETF	$77.41	$78.55	$71.17	+1.5%	-8.1%

Claude Mythos Preview is a general-purpose large language model that shows a major leap in capabilities over predecessors like Claude Opus 4.6, particularly in software engineering, reasoning, agentic tasks, and cybersecurity.

In internal and partner testing, the model autonomously:

• Identified thousands of high-severity and critical zero-day vulnerabilities (previously unknown flaws) in every major operating system, every major web browser, and numerous open-source projects.

• Turned many of those findings into working proof-of-concept exploits (e.g., full sandbox escapes, remote code execution, privilege escalation, and chained attacks).

• Achieved step-change benchmark results, such as 83.1% on CyberGym (vs. 66.6% for Opus 4.6), 100% pass@1 on a Cybench subset, and end-to-end success on private cyber ranges that no prior model completed.

Implication	Description	Why It Matters (Rationale)	43e
Defensive Product Enhancement	Use Mythos-level AI for autonomous vuln scanning, exploit chaining detection, and code hardening in EDR, SIEM, and cloud security tools.	Model finds zero-days and generates PoCs far faster than humans or legacy scanners.	New AI-powered “Mythos-class” scanning modules; faster patch recommendations; competitive edge for partners with early access.
Offensive Threat Amplification	Future public/similar models enable low-skill actors to launch advanced, autonomous attacks (e.g., custom zero-days overnight).	Drops the expertise and time required for exploits dramatically.	Must build stronger behavioral AI detection, sandboxing, and exploit-prevention layers; shorter detection windows expected.
Partnership & Access Advantage	Launch partners (CrowdStrike, Palo Alto, Cisco, etc.) get exclusive early access and collaboration.	Direct integration into security platforms and threat-intel sharing.	Accelerated R&D; co-developed defensive tools; potential revenue from AI-augmented services. Non-partners may lag.
Open-Source & Supply-Chain Security	Providers can scan and help patch foundational software (Linux kernel, browsers, FFmpeg, etc.) via Glasswing.	Thousands of previously unknown critical flaws in core dependencies.	Contribute to/fund open-source programs; integrate supply-chain risk scoring; position as “AI defenders of the internet.”
Market & Regulatory Pressure	Increased demand for AI-native security solutions; possible new compliance rules around AI-assisted vuln disclosure.	Governments and enterprises will require defenses against AI-powered threats.	Invest in AI talent/infrastructure; lobby for standards; prepare for audits on AI usage in security products.
Cost & Resource Implications	High token pricing + need for massive compute for agentic scanning.	Frontier models are expensive to run at scale.	Budget for API credits; optimize agentic workflows; explore on-prem or hybrid deployment once safeguards improve.
Ethical/Responsibility Shift	Providers become active participants in preemptive global hardening rather than just reactive responders.	Anthropic’s explicit goal: “put these capabilities to work for defensive purposes” before they proliferate.	Public reporting on patched vulns (90-day Glasswing updates); transparency on AI usage; align with responsible AI scaling.
Long-Term Industry Equilibrium	AI will eventually make software more secure overall (model-generated hardened code, automated patching).	Transitional risk is high, but net positive expected.	Pivot product roadmaps toward AI-augmented prevention and autonomous response; prepare for reduced reliance on signature-based detection.

It is available only in a tightly gated private preview via Project Glasswing, a defensive cybersecurity consortium. 

Launch partners include Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, Palo Alto Networks, and more than 40 additional organizations. 

For security software providers (antivirus/EDR vendors, firewall/endpoint firms, cloud security platforms, etc.), Claude Mythos Preview raises both the defensive opportunity and the offensive threat level.

Why it matters:

• Models can now autonomously find and exploit subtle, long-hidden vulnerabilities (some 16–27 years old) that survived millions of automated tests and human expert review 

• Defenders benefit by using Mythos Preview to scan their own products, customer environments, and critical open-source dependencies at superhuman speed and scale.

• Long-term equilibrium shifts are possible: (harder code, automated patching, faster incident response), but also increased attack volume and sophistication.

At least for the moment, investors seem unclear whether opportunity or risk is greater for incumbent suppliers of security products.